Local Binary Pattern of Maximum Gabor Response: A Novel Face Representation

In this paper, we proposed a novel face representation method: Local Binary Pattern of Maximum Gabor Response. This method is inspired by the HMAX model in biological vision. We first use multi-scales & multi-orientations Gabor filters to simulate the response of simple cells in visual cortex. For a given scale, we get the maximum responses maps from all orientations. Then LBP operator is used to describe the local features, and spatial pyramids are constructed to describe the spatial distribution information of face images. The proposed method is applied to face recognition, and the experiment results on ORL and FERET databases show that, compared with the state of art methods, the proposed algorithm still could achieve the comparable recognition rate with lower dimensional features.